Mythbusters: Is there evidence that promo games are troll-camped?

BeeCuz·Aug 4, 2016·3 votes·667 views

The scientific approach to this question is to create two possible outcomes, a hypothesis that expresses the researcher's query of interest and a null hypothesis that defines the rest of space. We then attempt to reject the null hypothesis.

Here then is that test performed against about 300 ranked matches https://s32.postimg.org/u6c72wec5/Untitled.png

Conclusion: The null hypothesis cannot be rejected with any significant level of confidence with the given data set.

dataset: https://docs.google.com/spreadsheets/d/11yGk3qnPmm63YtV3FkvovmKGrECks1FQnzVYUdPW8ZY/edit?usp=sharing

15 Comments

Pierce The VealAug 4, 20163 votes

I like that you tried to prove it empyrically, but there's simply too much anecdotal evidence to suggest that promos don't have anything rigged in them.

And honestly, odds are it is rigged in a way. Win three matches in a row (get in promos)? Matchmaking matches you against much stronger opponents to test your mmr.

Plus, from what I understand, you based this off of yourself? And testing players did so by either dodging the promos and/or trying to win promo then non promo over and over again? If you tried to set it up the easy way in any way whatsoever then by nature your tests are flawed. Best way to do it is to run some sort of double blind (not sure how to work this) or just work with the api to record what happens in all matches actoss thousands of players.

CoronaGulpAug 4, 20162 votes

Well said.

BeeCuzAug 4, 20161 votes

{quoted}

Plus, from what I understand, you based this off of yourself? And testing players did so by either dodging the promos and/or trying to win promo then non promo over and over again? If you tried to set it up the easy way in any way whatsoever then by nature your tests are flawed. Best way to do it is to run some sort of double blind (not sure how to work this) or just work with the api to record what happens in all matches actoss thousands of players.

Thank you for your opinion. I notice that it does not actually contain a test and an outcome.

Pierce The VealAug 4, 20161 votes

{quoted}

Thank you for your opinion. I notice that it does not actually contain a test and an outcome.

No problem. Well, part of what I said was somewhat opinionated, but the remainder definitely demolished your experiment. It's not that there wasn't a test or an outcome, no, the issue here is that your entire experiment says nothing except that the person who that data set was collected from wins more often in promos than in regular matches, which I highly doubt is the standard.

A data set of 11 promo games tells us nothing. There is no significane to data obtained from such a small set.

Thanks for trying anyway. I really appreciate it when people approach these things empyrically.

BloodyTomponAug 4, 20161 votes

I agree. I like the concept. I don't believe some of the assumptions hold true though (in particular, MMR is not likely to be stable if you're rising to promo's). The data set seems small and there is the question of is the individual representative for the population. Still interesting imo.

Pierce The VealAug 4, 20161 votes

{quoted}

I agree. I like the concept. I don't believe some of the assumptions hold true though (in particular, MMR is not likely to be stable if you're rising to promo's). The data set seems small and there is the question of is the individual representative for the population. Still interesting imo.

A data set with 11 promos is simply no data set at all.

BloodyTomponAug 4, 20161 votes

'The data set seems small' was a bit euphemistic haha :P

Pierce The VealAug 4, 20161 votes

{quoted}

'The data set seems small' was a bit euphemistic haha :P

Lol

520LegalizeRanchAug 4, 20161 votes

[{quoted}](name=Matsuda 13BREW,realm=NA,application-id=3ErqAdtq,discussion-id=4NNb3AJv,comment-id=,timestamp=2016-08-04T09:01:18.544+0000) I have played 30~ ranked games in the last week and have won nine. Prior to that, I have been on winning streaks leading up to three series' yet I have lost all but one of those 6-9 series games. I have since been demoted. 2/3 of my games consist of players being placed on my team who seem to like 0/500/0 KDAs and have no idea how to play at this level. I have carried many of these games as a jg but cannot win by myself.

BeeCuzAug 5, 20161 votes

{quoted}

The data set seems small and there is the question of is the individual representative for the population. Still interesting imo.

To answer your questions of dataset size and generalizability:

1). the necessary minimum size for this type of chi-square test is five points from the test set, and there are more than twice that, so the information is fairly accurate in saying it's not possible to reject the null hypothesis that game type (promo or regular) is independent of win/loss frequency.

2). The dataset is from one summoner, and so I can only relate as accurately as possible my own experience.

If you develop a wider dataset then please do share.

BloodyTomponAug 6, 20161 votes

If you develop a wider dataset then please do share.

Unfortunately I do not. The API probably has some way, but you might not want to invest the time learning that.

lolipopevelynnAug 4, 20162 votes

promos dont become rigged with trolls as long as you dont say you are in promos. once you say you are in promos you are pretty much shooting yourself in the foot.

Harvey RabbitAug 4, 20161 votes

I don't get the point you're trying to make, but I guess it has something to do with ranked games?

I don't think the Jamie and Adam play LoL....

Naag OneAug 4, 20161 votes

I don't recall a promo game that I have lost that was evenly contested. It has always been either a troll or some hardcore feeders in bot lane in the few promo games that I've managed to play.